RonPub -- Research Online Publishing
Not a member yet
199 research outputs found
Sort by
The Internet of Things as a Privacy-Aware Database Machine
Full text linkInstead of using a computer cluster with homogeneous nodes and very fast high bandwidth connections, we want to present the vision to use the Internet of Things (IoT) as a database machine. This is among others a key factor for smart (assistive) systems in apartments (AAL, ambient assisted living), offices (AAW, ambient assisted working), Smart Cities as well as factories (IIoT, Industry 4.0). It is important to massively distribute the calculation of analysis results on sensor nodes and other low-resource appliances in the environment, not only for reasons of performance, but also for reasons of privacy and protection of corporate knowledge. Thus, functions crucial for assistive systems, such as situation, activity, and intention recognition, are to be automatically transformed not only in database queries, but also in local nodes of lower performance. From a database-specific perspective, analysis operations on large quantities of distributed sensor data, currently based on classical big-data techniques and executed on large, homogeneously equipped parallel computers have to be automatically transformed to billions of processors with energy and capacity restrictions. In this visionary paper, we will focus on the database-specific perspective and the fundamental research questions in the underlying database theory
Hardware Accelerating the Optimization of Transaction Schedules via Quantum Annealing by Avoiding Blocking
Full text linkThe isolation property of database theory guarantees to avoid problems of not synchronized parallel execution of several transactions. In this paper we propose an algorithm for an optimal transaction schedule for the different cores of a multi-core CPU with minimal execution time ensuring the isolation property. Optimizing the transaction schedule is a combinatorial problem, which is ideal to be solved by quantum annealers as special form of quantum computers. In our contribution we show how to transform an instance of the transaction schedule problem into a formula that is accepted by quantum annealers including a proof of validity and optimality of the obtained result. Furthermore, we analyze the number of required qubits and the preprocessing time, and introduce an approach for caching formulas as result of preprocessing for the purpose of reducing the preprocessing time. In an experimental evaluation, the runtime on a quantum annealer outperforms the runtime of traditional algorithms to solve combinatorial problems like simulated annealing already for small problem sizes
Usability Testing and Evaluation of Multimedia E-Learning Management System in Higher Education: Criteria for Evaluation
Full text linkThis paper discusses the criteria for usability testing and evaluation of Multimedia e-Learning Management Systems (MEMS). This was achieved through an in-depth analysis and synthesis of literature and presentation of results of a practical application using the University of Zimbabwe MEMS, Towards Student-Centred Integration of Multimedia ELearning, TSIME. Firstly, a critical review, analysis and synthesis of usability testing and evaluation of MEMS was done. That was followed by an in-depth synthesis of the learning theories as the structural basis of MEMS. Major criteria were drawn from MEMS usability, design aspects, institutional dimensions, and learning theories. The derived criteria were merged with the generic usability heuristics producing sixteen TSIME Usability Heuristics, TSIMEUH. Heuristic Evaluation (HE) method was used to test TSIME. The evaluation was carried out for two weeks using three expert evaluators. Twenty-eight usability problems were identified from the study, ten of which were classified as requiring high priority intervention while the rest needed moderate to minimal priority intervention and were solved. The major criteria that emanated from the study were under motivation, ethics and navigation attributes. The key findings indicated that criteria drawn can empower learning and solidify a good learning environment through the use of MEMS. The criteria also embrace a consolidated learning pattern that can be used in global pandemics such as the COVID-19 era by both instructors and learners. The research concluded that learning is embedded in the historical, social and material context and should be improved through interaction and feedback that incorporate the learning theories
Due to COVID-19 the World's Activities Stopped, but not Research: Workshop on Very Large Internet of Things (VLIoT 2020)
Full text linkThe Very Large Internet of Things (VLIoT) workshop aims at discussing the solutions of problems arising especially for large-scale configurations. After continuously monitoring the global COVID-19 pandemic this year, the workshop changes the format the first time to an online event in order to overcome problems like travel restrictions. Besides missing face-to-face meetings the online format also has chances like an increased number of participants, less travel burdens and saving budget. Hence we received many high-quality submissions, from which we accepted 9 to be introduced in this editorial
Towards Knowledge Infusion for Robust and Transferable Machine Learning in IoT
Full text linkMachine learning (ML) applications in Internet of Things (IoT) scenarios face the issue that supervision signals, such as labeled data, are scarce and expensive to obtain. For example, it often requires a human to manually label events in a data stream by observing the same events in the real world. In addition, the performance of trained models usually depends on a specific context: (1) location, (2) time and (3) data quality. This context is not static in reality, making it hard to achieve robust and transferable machine learning for IoT systems in practice. In this paper, we address these challenges with an envisioned method that we name Knowledge Infusion. First, we present two past case studies in which we combined external knowledge with traditional data-driven machine learning in IoT scenarios to ease the supervision effort: (1) a weak-supervision approach for the IoT domain to auto-generate labels based on external knowledge (e.g., domain knowledge) encoded in simple labeling functions. Our evaluation for transport mode classification achieves a micro-F1 score of 80.2%, with only seven labeling functions, on par with a fully supervised model that relies on hand-labeled data. (2) We introduce guiding functions to Reinforcement Learning (RL) to guide the agents' decisions and experience. In initial experiments, our guided reinforcement learning achieves more than three times higher reward in the beginning of its training than an agent with no external knowledge. We use the lessons learned from these experiences to develop our vision of knowledge infusion. In knowledge infusion, we aim to automate the inclusion of knowledge from existing knowledge bases and domain experts to combine it with traditional data-driven machine learning techniques during setup/training phase, but also during the execution phase
Assuring Privacy-Preservation in Mining Medical Text Materials for COVID-19 Cases - A Natural Language Processing Perspective
Full text linkCurrently, there is a very large volume of Covid-19 related medical data that have been stored in cloud based systems and made available for studing the disease dynamics. without any privacy-preservation. In order to reduce possible privacy leakage and also accommodate massive medical reports with high efficiencies, we proposed a privacypreserving word embody-based text classification method for mining COVID-19 medical documents. It uses the recurrent neural network deep learning algorithm according to the identified internal hiding centralization pattern. In addition, a new model-fusion method is proposed for the continuous improvement of the system performance.The extensive numerical studies have demonstrated that the classifier of the proposed system has superior performance via integrating with the keywords extraction approach. Moreover, the advanced new model does not only accurately capture the keyword patterns but also effectively capture the analogical hierarchy structure of the pathology related datasets with lower computational complexity
Securing J2EE SOA Enterprise Applications with a Pattern-Based Approach
Full text linkSecurity is a key issue in SOA J2EE applications. The literature and a considerable number of frameworks address security issues for this type of enterprise application. However, there are two significant problems in this body of knowledge: (i) it is hard to find an architectural approach for dealing with security threats to SOA J2EE applications; and, (ii) technologies are constantly changing, making it is difficult to have an abstract view of the problems that are solved using specific technologies. The Core Security Patterns (CSP) catalogue solves both problems because it provides a comprehensive architectural solution to J2EE security issues and abstracts specific security technologies into security patterns. However, the CSP pattern catalogue is huge (more than 1,000 pages) and there are three significant challenges to understanding it completely: (i) the integration of the CSP security patterns and the Core J2EE Patterns (CJP) for the software architecture of SOA J2EE applications is not evident; (ii) the high abstraction level of the CSP patterns, in some cases, obscures the security problems that the patterns solve; and (iii) the implementation of the CSP patterns involves the configuration of complex security frameworks, adding a layer of complexity to securing a J2EE application using a pattern-based approach. To address these issues, we have developed a SOA multitier application based on the patterns described in the CJP catalogue, and we have secured it by implementing the patterns described in the CSP catalogue. This paper describes the work carried out during these developments. The main goal was to relate the CSP patterns with: (i) CJP patterns; (ii) the security concerns that the CSP patterns address; and (iii) the present security frameworks. As a result of this paper, we expect the inclusion of security elements in SOA enterprise applications to be easier for software architects and developers. Finally, four main conclusions can be drawn from our study: (i) security is an orthogonal aspect for SOA multitier development; (ii) implementation of security patterns relies heavily on security frameworks, with the configuration of security frameworks thus becoming one of the most complex issues when securing J2EE SOA multitier applications; (iii) no J2EE application servers are needed to deploy secure J2EE SOA enterprise applications; and (iv) whether or not applications servers are used, security-related implementations are closely tied to the application container and frameworks used for SOA implementation
On Distributed SPARQL Query Processing Using Triangles of RDF Triples
Full text linkKnowledge Graphs are providing valuable functionalities, such as data integration and reasoning, to an increasing number of applications in all kinds of companies. These applications partly depend on the efficiency of a Knowledge Graph management system which is often based on the RDF data model and queried with SPARQL. In this context, query performance is preponderant and relies on an optimizer that usually makes an intensive usage of a large set of indexes. Generally, these indexes correspond to different re-orderings of the subject, predicate and object of a triple pattern. In this work, we present a novel approach that considers indexes formed by a frequently encountered basic graph pattern: triangle of triples. We propose dedicated data structures to store these triangles, provide distributed algorithms to discover and materialize them, including inferred triangles, and detail query optimization techniques, including a data partitioning approach for bias data. We provide an implementation that runs on top of Apache Spark and experiment on two real-world RDF data sets. This evaluation emphasizes the performance boost (up to 40x on query processing) that one can obtain by using our approach when facing triangles of triples
Information-Centric Semantic Web of Things
Full text linkIn the Semantic Web of Things (SWoT) paradigm, a plethora of micro-devices permeates an environment. Storage and information processing are decentralized: each component conveys and even processes a (very) small amount of annotated metadata. In this perspective, the node-centric Internet networking model is inadequate. This paper presents a framework for resource discovery in semantic-enhanced pervasive environments leveraging an information-centric networking approach. Information gathered through different Internet of Things (IoT) technologies can be exploited by both ubiquitous and Web-based semantic-aware applications through a uniform set of operations. Experimental results and a case study support sustainability and effectiveness of the proposal
Ten Ways of Leveraging Ontologies for Rapid Natural Language Processing Customization for Multiple Use Cases in Disjoint Domains
Full text linkWith the ever-growing adoption of AI technologies by large enterprises, purely data-driven approaches have dominated the field in the recent years. For a single use case, a development process looks simple: agreeing on an annotation schema, labeling the data, and training the models. As the number of use cases and their complexity increases, the development teams face issues with collective governance of the models, scalability and reusablity of data and models. These issues are widely addressed on the engineering side, but not so much on the knowledge side. Ontologies have been a well-researched approach for capturing knowledge and can be used to augment a data-driven methodology. In this paper, we discuss 10 ways of leveraging ontologies for Natural Language Processing (NLP) and its applications. We use ontologies for rapid customization of a NLP pipeline, ontologyrelated standards to power a rule engine and provide standard output format. We also discuss various use cases for medical, enterprise, financial, legal, and security domains, centered around three NLP-based applications: semantic search, question answering and natural language querying